Hybrid Architectures and Multiscale Methods – HAMM

This research project aims at the development, analysis and software
implementation of mathematical models for multiscale applications on hybrid
architectures. Large scale multiscale applications are indeed within reach
with the emerging computing infrastructures, but they require accurate and
robust multiscale numerical methods that take into account these new
architectures. This is very challenging as current software were not designed
for these methods and architectures. As a matter of fact, scientists and
engineers need to manage (i) the complexity of the underlying
multiscale models, usually expressed in terms of a partial differential
equation (PDE) system completed with algebraic closure laws, (ii) the
complexity of numerical methods used to solve the PDE systems, and finally
(iii) the complexity of the low level computer science services
required to have efficient software on modern hybrid hardware (e.g. multicore
CPU/GPU). Robust and effective multiscale methods as well as advanced
programming techniques need to be combined to fully benefit from massively
hybrid parallel architectures.

Even though this project revolves around applied mathematics and multiscale
numerical methods, it is truly multidisciplinary and critically requires the
expertise of specialists of the applications, computer scientists and computer
designers. That's why we create a strong consortium : (i) CEA and IFP
leading the large scale multiscale application specifications, (ii) UJF
leading the multiscale numerical methods development and efficient
implementation on hybrid architectures and finally (iii) BULL leading the
benchmarking activities and ultimately the sizing of future super computing
infrastructure according to nowadays constraints such as energy costs savings.